Also available at http://amc-journal.eu ISSN 1855-3966 (printed edn.), ISSN 1855-3974 (electronic edn.) ARS MATHEMATICA CONTEMPORANEA 14 (2018) 329–344 A note on the thickness of some complete bipartite graphs Siwei Hu , Yichao Chen ∗ Department of Mathematics, Hunan University, 410082 Changsha, China Received 22 November 2016, accepted 17 July 2017, published online 19 September 2017 Abstract The thickness of a graph is the minimum number of planar subgraphs into which the graph can be decomposed. Determining the thickness for the complete bipartite graph is an unsolved problem in graph theory for over fifty years. Using a new planar decomposition for K4k−4;4k(k ≥ 4), we obtain the thickness of the complete bipartite graph Kn;n+4; for n ≥ 1: Keywords: Planar graph, thickness, complete bipartite graph. Math. Subj. Class.: 05C10 1 Introduction In this paper, all graphs are simple. A graph G is denoted by G = (V; E) where V (G) is the vertex set and E(G) is the edge set. A complete graph is a graph in which any two vertices are adjacent. A complete graph on n vertices is denoted by Kn: A complete bipartite graph is a graph whose vertex set can be partitioned into 2 parts, such that every edge has its ends in different parts and every two vertices in different parts are adjacent. We use Kp1;p2 to th denote a complete bipartite graph in which the i part contains pi vertices, for i = 1; 2: The thickness t(G) of a graph G is the minimum number of planar subgraphs into which G can be decomposed [15]. It is a classical topological parameter of a graph and has many applications, for instance, to graph drawing [3] and VLSI design [1]. Since deciding the thickness of a graph is NP-hard [11], it is very difficult to get the exact number of thickness for arbitrary graphs. Battle, Harary and Kodama [4] in 1962 and Tutte [14] in 1963 independently showed that the thickness of K9 and K10 equals 3. Beineke and Harary [5] determined the thickness of complete graph Kn for n 6≡ 4 (mod 6) in 1965, the remaining case was solved in 1976, independently by V.B. Alekseev and V.S. Gonchakov [2] and by J.M. Vasak [13]. ∗Corresponding author. E-mail addresses: [email protected] (Siwei Hu), [email protected] (Yichao Chen) cb This work is licensed under http://creativecommons.org/licenses/by/3.0/ 330 Ars Math. Contemp. 14 (2018) 329–344 For complete bipartite graphs, the problem has not been entirely solved yet. By con- structing a planar decomposition of Km;n when m is even, Beineke, Harary and Moon [6] determined the thickness of Km;n for most values of m; n in 1964. Theorem 1.1. [6] For m ≤ n, the thickness of the complete bipartite graph Km;n is mn t(K ) = ; (1.1) m;n 2(m + n − 2) except possibly when m and n are both odd and there exists an integer k satisfying n = j 2k(m−2) k (m−2k) . We recall that the thickness of Kn;n is also obtained in 1968 by Isao and Ozaki [10] independently. The following open problem is adapted from [8] by Gross and Harary. Problem 1.2. [See Problem 4.1 of [8]] Find the thickness of Km;n for all m; n: Beineke, Harary and Moon [6] also pointed out that the smallest complete bipartite graph whose thickness is unknown is K17;21: From Euler’s Formula, the thickness of K17;21 is at least 5: From Theorem 1.1, we need to determine the thickness of Km;n for odd m; n. Since the difference between the two odd numbers is even, we only need to determine the thickness of Kn;n+2k for odd n and k ≥ 0: In this paper, we start to calculate the thickness of Kn;n+2k for some small values of k. Indeed, we determine the thickness of Kn;n+4: Theorem 1.3. The thickness of Kn;n+4 is ( 1; if n ≤ 2 t(Kn;n+4) = n+3 4 ; otherwise. The following corollary follows from Theorem 1.3. Corollary 1.4. The thickness of K17;21 is 5. We may refer the reader to [7, 12, 16] for more background on graph thickness. 2 The thickness of Kn;n+4 To begin with, we define two special graphs called the pattern graph and the kth-order nest graph. Then, we prove a new planar decomposition of K4k−4;4k: Finally, we prove the thickness of K4k−3;4k+1 and Kn;n+4: 2.1 The pattern graph Let U = fu1; u2g and Xn be a set of n vertices. A graph is said to be a pattern graph of order n + 2; denoted by G[u1Xnu2]; if it can be constructed by the following two steps. 1. Arrange the n vertices in a row, and put vertices u1; u2 on the above and below of n vertices, respectively. 2. Join both u1 and u2 to the n vertices using straight lines. S. Hu and Y. Chen: A note on the thickness of some complete bipartite graphs 331 From the definition above, the pattern graph is a planar straight-line graph. Figure1 illus- trates the pattern graph G[u1Xnu2]. Remark 2.1. Unless explicitly mentioned, we always join vertices using straight lines in the drawings of the following proofs. u1 u2 Figure 1: The pattern graph G[u1Xnu2]: 2.2 The kth-order nest graph Let Uk = fui1 ; ui2 ; : : : ; uik g;Vk = fvj1 ; vj2 ; : : : ; vjk g and W2k+2 = fwl1 ; wl2 ;:::; th wl2k+2 g; we define a k -order nest graph G[Uk;Vk;W2k+2] as follows: 1. Arrange 2k + 2 vertices wl1 ; wl2 ; : : : ; wl2k+2 in a row. 2. For 1 ≤ m ≤ k; place vertices uim and vjm on the above and below of the row, respectively, and join them to wl1 ; wl2m ; wl2m+1 ; wl2m+2 : Figure2 illustrates a third-order nest graph G[U3;V3;W8]; where U3 = fu1; u2; u3g; V3 = fv1; v2; v3g and W8 = fw1; w2; : : : ; w8g. u3 u2 u1 w1 w2 w3 w4 w5 w6 w7 w8 v1 v2 v3 Figure 2: The third-order nest graph G[U3;V3;W8]. 2.3 A new planar decomposition of K4k−3;4k+1, for k ≥ 4 In this subsection, we shall construct a planar decomposition for the complete bipartite graph K4k−3;4k+1 with k planar subgraphs G1;G2;:::;Gk: Suppose that the vertex parti- tion of K4k−3;4k+1 is (X; Y ), where X =fx1; x2; : : : ; x4k−3g;Y =fy0; y1; y2; : : : ; y4kg: 332 Ars Math. Contemp. 14 (2018) 329–344 2.3.1 The planar decomposition for K4k−4;4k Let the vertex partition of K4k−4;4k be (X1;Y1), where X1 = fx1; x2; : : : ; x4k−4g;Y1 = fy0; y1; : : : ; y4k−1g. In this subsection, all subscripts in yj are taken mod 4k: 1. In the graph Gi (1 ≤ i ≤ k), we arrange 4k vertices in a row, and divide the 4k vertices into two subsets L2k and R2k such that each subset contains 2k vertices according to the following steps. 2. In the graph Gi (1 ≤ i ≤ k − 1), we choose four vertices x4i−3; x4i−2; x4i−1; x4i from X1 and construct two pattern graphs G[x4i−3L2kx4i−1] and G[x4i−2R2k x4i]. Then we join both x4i−3 and x4i−1 to the first vertex and the last vertex in R2k: Fi- nally, we label the vertices in L2k and R2k as y1; y3; y5; : : : ; y4k−1 and y2i+6; y2i+8; y2i+10; : : : ; y2i+4k+4 in turn, respectively. 3. In the graph Gk, we label the vertices in L2k and R2k as y1; y3; y5; : : : y4k−1 and th y2; y4; : : : ; y4k−2; y0; respectively. First, we construct a (k − 1) -order nest graph G[Uk−1;Vk−1;W2k]; where Uk−1 = fx2; x6; x10; : : : ; x4k−6g ;Vk−1 = fx4; x8; x12; : : : ; x4k−4; g and W2k = fy1; y3; y5; : : : ; y4k−1g. We join x4i−3 to y2i and y2i+2, for 1 ≤ i ≤ k − 1: Second, we construct a union of paths, if k is even, we join x4i−1 to y2i+2k and y2i+2+2k, for 1 ≤ i ≤ k − 1; otherwise k is odd, we join x4i−1 to y2i+2k−2 and y2i+2k, for 1 ≤ i ≤ k − 1: 4. In each graph Gj (1 ≤ j ≤ k − 1), we put x4i−2; x4i in the quadrangle x4j−3y4j+1 x4j−1y4j+3; and join them to y4j+1 and y4j+3; for 1 ≤ i < j: We put the vertices x4i−2; x4i in the quadrangle x4j−3y4j−1x4j−1y4j+1, and join both x4i−2 and x4i to y4j−1 and y4j+1; for j < i ≤ k − 1: Next, we put x4i−3 in the quadrangle x4j−2y4j−2i+4x4jy4j−2i+6; and join x4i−3 to y4j−2i+4; y4j−2i+6, for 1 ≤ i < j: We put x4i−3 in the quadrangle x4j−2y4j−2i+4kx4jy4j−2i+4k+2, and join x4i−3 to y4j−2i+4k; y4j−2i+4k+2, for j < i ≤ k − 1: For each i (1 ≤ i ≤ k − 1), we define a set Mi = fi + 1; i + 2; : : : ; i + k − 2g: Suppose that m 2 Mi, if m ≤ k − 1, we let j = m; otherwise, j = m − k + 1. k−4 (i) k is even. If i+1 ≤ m ≤ i+ 2 , we put x4i−1 in the quadrangle x4j−2y4m−2i+4 k−4 x4jy4m−2i+6; and join x4i−1 to y4m−2i+4; y4m−2i+6.